Control-Oriented Modeling of Pipe Flow in Gas Processing Facilities

TL;DR

This paper develops linearized, control-oriented models of pipe flow in gas processing facilities, integrating nonlinear physics-based equations into network models suitable for feedback control design and validated with real operational data.

Contribution

It introduces a systematic approach to generate and validate simplified, linearized pipe network models from complex physics-based equations for control applications.

Findings

Models accurately describe pressures, flows, and temperatures.

Interconnection method aligns with Mason's Gain Formula.

Validated models show acceptable divergence from physics-based data.

Abstract

Pipe flow models are developed with a focus on their eventual use for feedback control design at the process control level, as opposed to the unit level, in gas processing facilities. Accordingly, linearized facility-scale models are generated to describe pressures, mass flows and temperatures based on sets of nonlinear partial differential equations from fluid dynamics and thermodynamics together with constraints associated with their interconnection. As part of the treatment, the divergence of these simplified models from physics is assessed, since robustness to these errors will be an objective for the eventual control system. The approach commences with a thorough analysis of pipe flow models and then proceeds to study their automated interconnection into network models, which subsume the algebraic constraints of bond graph or standard fluid modeling. The models are validated and their errors quantified by referring them to operational data from a commercial gas compressor test facility. For linear time-invariant models, the interconnection method to generate network models is shown to coincide with automation of Mason's Gain {Formula}. These pipe network models based on engineering data are the first part of the development of general facility process control tools.